Multiphase turbulence in bubbly flows: RANS simulations

The ability of a two-fluid Eulerian–Eulerian computational multiphase fluid dynamic model to predict bubbly air–water flows is studied. Upward and downward pipe flows are considered and a database of 19 experiments from 6 different literature sources is used to assess the accuracy of the model, with the aim of evaluating our ability to predict these kinds of flows and to contribute to ongoing efforts to develop more advanced simulation tools. The particular focus in the work described is on the prediction of multiphase turbulence due to its relevance in the modelling of bubbly flows in general, including bubble coalescence and break-up, and boiling at a wall. Overall, a satisfactory accuracy is obtained in the prediction of liquid velocities and void fraction distributions in all conditions, including upward and downward flows, and wall-peaked and core-peaked void profiles, when values of the bubble diameter are specified from experimental data. Due to its importance for the correct prediction of the turbulence level in these flows, a bubble-induced turbulence model is introduced, starting from an existing formulation. Source terms due to drag are included in the turbulence kinetic energy and the turbulence energy dissipation rate equations of the k-ε turbulence model, and optimisation of the turbulence source gives velocity fluctuation predictions in agreement with data. After comparisons with data, improvement in the predictions of other turbulence models is also demonstrated, with a Reynolds stress formulation based on the SSG (Speziale et al., 1991) pressure–strain model and the same bubble-induced turbulence model accurately predicting the two-phase flows and the anisotropy of the turbulence field. The same database is also exploited to evaluate different drag models and the advantages of including the effect of the bubble aspect ratio. Following experimental evidence, the model of Tomiyama et al. (2002) is used which assumes that the bubble shape is closer to spherical near a wall and employs a correlation to calculate the aspect ratio. An increase in the drag coefficient due to the higher aspect ratio increases the accuracy of calculated velocity profiles in the near-wall region, even if additional validation is still required due to the possible loss of accuracy in the pipe centre